Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 7 Dec 2023 17:02:57 -0800
From:      Maxim Sobolev <sobomax@freebsd.org>
To:        Bakul Shah <bakul@iitbombay.org>
Cc:        Warner Losh <imp@bsdimp.com>, Tomoaki AOKI <junchoon@dec.sakura.ne.jp>,  FreeBSD Current <freebsd-current@freebsd.org>
Subject:   Re: nvme timeout issues with hardware and bhyve vm's
Message-ID:  <CAH7qZfuC8WHUpSvsT2tQo-9txcWkTg84GXbGHR5uBXtQaFw1aQ@mail.gmail.com>
In-Reply-To: <10FD2FC6-1F39-4F7D-8BA8-976ADC0AE37A@iitbombay.org>
References:  <90d3e532-8ea7-4eea-8e31-8c363285a156@nomadlogic.org> <CANCZdfrQTd3F-j81HsamUCJG4DyUk_-yPOtbZY4Q926_ihatsQ@mail.gmail.com> <0ad493d5-1c1e-4370-977a-118f46ebd677@nomadlogic.org> <CANCZdfrwzmZ=iHj_vm2nsi72ceRQ81KY5DjiuML3udEaWTBanA@mail.gmail.com> <0c4f8149-89dd-4635-a5ed-4766fffd2553@nomadlogic.org> <CANCZdfpgw_sm4couYx9%2Bcgp-q_2jmPC2Q7TSeD9Yb3VYoiDQhQ@mail.gmail.com> <ec08484d-b49f-4aa3-adf4-b96570083b9c@nomadlogic.org> <20231208080929.cfd9fca421fea81d89d2380b@dec.sakura.ne.jp> <CANCZdfpN=GwJQT%2BrK=TMUj6niajw-0C=957gV655s2FqJ79nKw@mail.gmail.com> <10FD2FC6-1F39-4F7D-8BA8-976ADC0AE37A@iitbombay.org>

next in thread | previous in thread | raw e-mail | index | archive | help
--000000000000c61723060bf525a2
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

How quickly it heats up depends on lots of factors. Usually those devices
burn some 3-7 watts per stick at 100% load, so maybe this would give you
some idea. At least some of them support several toggleable performance
modes, which use throttling internally to limit power consumption to a
certain level (man nvmecontril). It helped me recently to make a system
stable, which otherwise would hang with timeout after reaching 70-75C until
I got the chance to take it apart and attach a heatsinks to the nvmes. Once
the temperature dropped to <=3D 50C the drives become 100% stable.

-Max

On Thu, Dec 7, 2023, 4:07=E2=80=AFPM Bakul Shah <bakul@iitbombay.org> wrote=
:

> On Dec 7, 2023, at 3:59=E2=80=AFPM, Warner Losh <imp@bsdimp.com> wrote:
> >
> >
> >  *Overheating caused hang of NVMe controller or PCI bridge on SSD, or
> >
> > Yes. Most drive's firmware when it overheats resets. There might be
> something
> > that the pci code can do when this happens to retrain the link,
> reprogram the
> > config registers, etc.
>
> How quickly can the device heat up? Can it be queried frequently
> enough act before it overheats by throttling io?
>
>
>
>
>

--000000000000c61723060bf525a2
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"auto">How quickly it heats up depends on lots of factors. Usual=
ly those devices burn some 3-7 watts per stick at 100% load, so maybe this =
would give you some idea. At least some of them support several toggleable =
performance modes, which use throttling internally to limit power consumpti=
on to a certain level (man nvmecontril). It helped me recently to make a sy=
stem stable, which otherwise would hang with timeout after reaching 70-75C =
until I got the chance to take it apart and attach a heatsinks to the nvmes=
. Once the temperature dropped to &lt;=3D 50C the drives become 100% stable=
.<div dir=3D"auto"><br></div><div dir=3D"auto">-Max</div></div><br><div cla=
ss=3D"gmail_quote"><div dir=3D"ltr" class=3D"gmail_attr">On Thu, Dec 7, 202=
3, 4:07=E2=80=AFPM Bakul Shah &lt;<a href=3D"mailto:bakul@iitbombay.org">ba=
kul@iitbombay.org</a>&gt; wrote:<br></div><blockquote class=3D"gmail_quote"=
 style=3D"margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On=
 Dec 7, 2023, at 3:59=E2=80=AFPM, Warner Losh &lt;<a href=3D"mailto:imp@bsd=
imp.com" target=3D"_blank" rel=3D"noreferrer">imp@bsdimp.com</a>&gt; wrote:=
<br>
&gt; <br>
&gt; <br>
&gt;=C2=A0 *Overheating caused hang of NVMe controller or PCI bridge on SSD=
, or<br>
&gt; <br>
&gt; Yes. Most drive&#39;s firmware when it overheats resets. There might b=
e something<br>
&gt; that the pci code can do when this happens to retrain the link, reprog=
ram the<br>
&gt; config registers, etc.<br>
<br>
How quickly can the device heat up? Can it be queried frequently<br>
enough act before it overheats by throttling io?<br>
<br>
<br>
<br>
<br>
</blockquote></div>

--000000000000c61723060bf525a2--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAH7qZfuC8WHUpSvsT2tQo-9txcWkTg84GXbGHR5uBXtQaFw1aQ>